BYOK & models
Molar is provider-agnostic. You can run it entirely on Molar's hosted models, or bring your own key (BYOK) and route your LLM calls through your own provider account. No single vendor's key is required end to end.
Hosted models
By default, Molar routes model calls through its LLM gateway to a managed pool (Azure, DeepSeek, NVIDIA Nemotron, and others). Pick a model in Settings → Models. Hosted usage counts toward your organization's hosted-LLM spend cap in the dashboard.
Bring your own key (BYOK)
Add a provider key under Settings → Models → Provider keys. Once set, your organization's LLM calls are routed through your key instead of the hosted pool:
- Your keys are stored encrypted and never returned in API responses or logs.
- BYOK calls bypass the hosted gateway, so they are not subject to the hosted-LLM spend cap — when your org hits the hosted cap, BYOK calls keep working.
- You are responsible for usage and charges on your provider account.
Choosing a routing strategy
| Goal | Setting |
|---|---|
| Simplest — no provider account | Hosted models |
| Cost control / data residency | BYOK with your provider |
| Mix — cheap routine calls, your key for sensitive ones | Hosted + BYOK per surface |
Spend caps
Hosted-LLM spend is capped per organization to protect shared infrastructure. The cap blocks hosted calls only; BYOK remains available. See Billing → Usage on app.molar.it.
How to add a BYOK key
- Sign in at app.molar.it
- Open Settings → Models → Provider keys
- Choose your provider (OpenAI, Anthropic, Azure OpenAI, or any OpenAI-compatible endpoint)
- Paste the key — Molar encrypts it at rest and routes only your org's calls through it
- Pick which surfaces use BYOK vs hosted models in Settings → Models
Test routing with a Cartographer crawl or Trace Debugger message. BYOK calls appear on your provider bill, not Molar's hosted-LLM meter.
See also LLM gateway for routing details.